See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/283278961

Slow sound in a duct, effective transonic flows,

and analog black holes
ARTICLE in PHYSICAL REVIEW D OCTOBER 2015
Impact Factor: 4.64 DOI: 10.1103/PhysRevD.92.081503

READS

51

5 AUTHORS, INCLUDING:
Yves Aurgan

Vincent Pagneux

Universit du Maine

French National Centre for Scientific Research

71 PUBLICATIONS 863 CITATIONS

132 PUBLICATIONS 1,044 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Vincent Pagneux

Retrieved on: 18 November 2015

RAPID COMMUNICATIONS

PHYSICAL REVIEW D 92, 081503(R) (2015)

Slow sound in a duct, effective transonic flows, and analog black holes
Yves Aurgan,1,* Pierre Fromholz,2, Florent Michel,3, Vincent Pagneux,1, and Renaud Parentani3,
1

Laboratoire dAcoustique, Universit du Maine, UMR CNRS 6613, Avenue O Messiaen,

72085 Le Mans Cedex 9, France
2
Dpartement de Physique, ENS, 24 rue Lhomond, 75005 Paris, France
3
Laboratoire de Physique Thorique, CNRS, Universit Paris-Sud, Universit Paris-Saclay,
91405 Orsay, France
(Received 16 March 2015; published 21 October 2015)
We propose a new system suitable for studying analog gravity effects, consisting of a gas flowing in a
duct with a compliant wall. Effective transonic flows are obtained from uniform, low-Mach-number flows
through the reduction of the one-dimensional speed of sound induced by the wall compliance. We show
that the modified equation for linear perturbations can be written in a Hamiltonian form. We perform a
one-dimensional reduction consistent with the canonical formulation, and deduce the analog metric along
with the first dispersive term. In a weak dispersive regime, the spectrum emitted from a sonic horizon is
numerically shown to be Planckian, and with a temperature fixed by the analog surface gravity.
DOI: 10.1103/PhysRevD.92.081503

PACS numbers: 04.62.+v, 04.70.Dy, 43.20.+g, 43.20.Wd

I. INTRODUCTION
Engineering flows that are transonic and regular offers
the possibility to test well-known predictions concerning
astrophysical black holes [1]. Of particular interest is
Hawkings discovery that black holes should spontaneously emit a steady thermal flux [2]. Although this effect
was originally phrased in the context of quantum relativistic fields, it rests on the anomalous mode mixing
occurring near the black-hole horizon [3]. This mixing,
which is stationary and conserves the wave energy, is called
anomalous because it leads to a mode amplification and
involves negative energy waves. Because of the precise
analogy between the equation governing sound propagation and that used by Hawking, these key elements are
recovered in a stationary transonic flow. Indeed, in the
acoustic approximation, for long wavelengths, the mode
mixing possesses the main properties of the one responsible
for the Hawking effect [4,5].
To complete the comparison, one should take into
account the dispersive properties of sound waves, which
have no counterpart in general relativity. (Note, however,
that dispersive terms appear in certain theories of modified
gravity where Lorentz invariance is broken [68].)
Analytical and numerical studies have established that
the correspondence is quantitatively preserved provided
the two relevant scales are well separated [4,5,912],
namely, when the dispersive length is sufficiently smaller
than the typical length scale associated with the inhomogeneity of the flow (which then plays the role of the
inverse surface gravity of the black hole). Therefore, there
is no conceptual obstacle preventing the testing of the
*

yves.auregan@univlemans.fr
pierre.fromholz@ens.fr
florent.michel@th.upsud.fr

vincent.pagneux@univlemans.fr

renaud.parentani@th.upsud.fr

1550-7998=2015=92(8)=081503(5)

Hawking prediction by observing the mode mixing across

a sonic horizon. In practice, the difficulty is finding
appropriate setups. Many have been proposed, involving
for instance ultracold atomic clouds [13], surface waves in
flumes [14], and light in nonlinear media [15]. Recently,
the first experiments have been carried out [1619].
In this paper, we propose a new framework which is a
variant of the original one [1,4]. Its main interest resides in
obtaining a large reduction of the low-frequency onedimensional sound speed in a duct so that a stationary
flow with a uniform low Mach number M possesses a
horizon. In realistic settings, M could be close to 0.3.
The reduction of the effective sound speed is achieved
by means of a compliant wall composed of thin tubes
which modify the upper boundary condition, see Fig. 1.

FIG. 1 (color online). Schematic drawing of the configuration.

At y 0, the wall is rigid. At y 1, the compliant wall is made
of a succession of tiny tubes of height bx. As explained in the
text, the effective sound speed is a decreasing function of b. In the
present profile bx and for a given uniform flow with Mach
number M < 1, an effective supersonic region can be created in
the right region.

081503-1

2015 American Physical Society

RAPID COMMUNICATIONS

YVES AURGAN et al.

PHYSICAL REVIEW D 92, 081503(R) (2015)

Interestingly, for near-critical flows, the wave equation

follows from a well-defined action principle. From this,
we derive the acoustic black-hole metric in the longwavelength regime, and a conserved norm which establishes the anomalous character of the mode mixing. At the
end of this paper we briefly present the practical advantages
of this proposal.
II. THE MODEL
We consider the propagation of sound waves in a
two-dimensional horizontal channel of uniform height H.
x denotes the Cartesian horizontal coordinate, y the vertical
one, and t the time. We assume the flow of air is uniform,
with a horizontal velocity U 0. Denoting by c0 the sound
speed, 0 the air density, v the velocity perturbation, and p
the pressure perturbation, the time evolution is given by
 


c1
0 Dt p 0 v ;

0 Dt v  p ; 1

where Dt t U0 x is the convective derivative. We

define dimensionless quantities as x x =H, y y =H,
t t c0 =H, v v =c0 , p p =0 c20 , and M U0 =c0 .
The potential gives the velocity by v , and the
pressure by p Dt . It obeys
D2t 2x 2y 0:

At the lower wall, the impenetrability condition is simply

y t; x; y 0 0, see Fig. 1. At the upper wall y 1,
the continuity of the displacement and pressure gives rise to
a nonlocal expression in time, see [2022] for details.
However, for near-critical flows and small frequencies, it
can be written as
y Dt bxDt 0 at y 1;

which is second order in t .

For a homogeneous stationary flow, we can look for
solutions of the form k coshk yeikxk t . Equation (2)
and the two boundary conditions respectively give
2k k2 2k

and k tanhk b2k ;

where k k Mk is the frequency in the comoving

frame. At low wave number, the dispersion relation reads
2k c2S bk2 k4 =2b Ok6 ;

where
c2S b

1
;
1b

2b

31 b3
:
b2

One sees the important effect of the boundary condition of

Eq. (3): the low-frequency group velocity with respect to the

fluid ( k ) is reduced by the compliant wall. One also

sees that the dispersive length 1=b given by the quartic term
vanishes in the limit b 0.
To obtain flows crossing the effective sound speed,
we make b vary with x, see Fig. 1. We call b1 > b2 its
asymptotic values, and db its typical variation length. We
choose the following form for bx:


b b2 b2 b1
x
bx 1
:
7

tanh
db
2
2
We then adjust the flow speed M to obtain
1
1
cS b1 p < jMj < p cS b2 :
1 b1
1 b2

Since the background flow is stationary, we shall work with

(complex) stationary waves,
Z
dt it
~ x; y
e t; x; y:
9
2
III. WAVE EQUATION
Since the height H of the duct is much smaller than
typical longitudinal wavelengths, we expect that stationary
waves obey an effective one-dimensional equation in x, as
is the case in elongated atomic Bose condensates [23] and
in flumes [14,24]. To obtain such a reduction is nontrivial,
as the x dependence of bx prevents us from factorizing
out a y-dependent factor. To proceed, and to make contact
with the works mentioned above, it is useful to exploit the
fact that Eq. (2) and Eq. (3) can be derived from the
following action:
Z
Z
Z 1
1
S
dt dx
dyL
2
0
L Dt 2 2 y 1bxDt 2 :

10

We notice that the nontrivial condition of Eq. (3) is

incorporated by the above boundary term. Introducing
the conjugate momentum t; x; y L= t one
obtains the Hamiltonian H by the usual Legendre transformation. In addition, as in [4,24], the conserved inner
product j is not positive definite, and has the KleinGordon form
Z Z 1
1 ; 2 i
dx
dy 1 2 1 2 ; 11

where 1 ; 2 are two complex solutions of Eqs. (2), (3), and

1 ; 2 their associated momenta. The inner product is
conserved by virtue of Hamiltons equations. As for sound
waves in other media, it identically vanishes for all real
solutions. However it provides key information when
studying stationary modes, namely, the sign of their norm
; . Indeed, as we shall see, for a fixed > 0, there

081503-2

RAPID COMMUNICATIONS

SLOW SOUND IN A DUCT, EFFECTIVE TRANSONIC

PHYSICAL REVIEW D 92, 081503(R) (2015)

will be both positive and negative norm modes. When the

flow is stationary, for any complex solution , the wave
energy is conserved and related to Eq. (11) by
H2Re ; i t :

12

Moreover, when the flow is also asymptotically homogeneous, for every asymptotic plane wave k
coshk yeikxk t , the sign of H is that of k k . This
relation will allow us to identify the negative energy waves
without ambiguity.
We can now proceed following the hydrodynamic treatment of [24]. As a first step, it is useful to derive a (1 1)dimensional equation from which an effective space-time
metric can be read out. When the (adimensional) wavelength in the x direction is much larger than 1, we can
assume that 2y is independent of y. As y 0 at y 0,
we write the field as
x; y; t x; t y2 x; t:

13

Plugging this into the action Eq. (10) and varying it with
respect to and , we get two coupled equations.
2O
4 0, where O
n
Combining them, we obtain O
is an nth-order operator in t and x . The quadratic term is
2 F x , where
O

F x

c2S bx M2


;

14

and c2S bx is given in Eq. (6). Up to a conformal factor,

we obtain the dAlembert equation in a two-dimensional
space-time with metric g F . This metric has a Killing
horizon where c2S x M2 vanishes [25]. This correspondence with gravity relates the anomalous scattering
described below to the Hawking effect.
Contrary to what happens for sound waves in atomic
Bose-Einstein condensate, or water waves in the incom 4 also contains third and fourth derivatives
pressible limit, O
in time. This prevents us from applying the standard treatment on the sole field . However, the set of two coupled
equations on ; is Hamiltonian and can be used to study
the scattering. Alternatively, one can work with the original
model in 2 1 dimensions based on Eq. (10). We performed
numerical simulations with both models and found similar
results.
IV. ANOMALOUS MODE MIXING
Since stationary waves with different frequencies do
not mix, the scattering only concerns the discrete set of
modes with the same . To characterize it, we identify its
dimensionality and the norms of the various asymptotic
modes for x .
Figure 2 shows the dispersion relation and the roots at
fixed in a subsonic flow for M > 0. For this sign of M,

FIG. 2 (color online). Dispersion relation versus k in a

homogeneous subsonic flow. The blue solid curves show the
roots with positive comoving frequency , and the red dashed
ones show those with < 0. The dotted black line represents
0.4, and the dot-dashed one shows the value max of at
which the two roots with k < 0 merge. The parameters are b 1,
Mp
0.4, and the effective sound velocity cS is equal to
1= 2 0.71.

the flow associated with Eq. (7) passes from supersonic to

subsonic along the direction of the stream. It thus corresponds to a white-hole flow, like those studied in
[17,24,26].1 For definiteness, we discuss only the case
> 0. The same results are directly applicable to < 0
after complex conjugation. In the subsonic region, on the
right of the horizon, there exists a critical frequency max
(close to 0.42 in Fig. 2) at which two roots merge. For
< max , the dispersion relation has four real roots.2
in
Following [24], we call their wave vectors kout
, k ,
co;out
out
k , and k . The corresponding asymptotic modes
co;out
in

are, respectively, out
, and out
, ,
. They are
characterized by three important properties:
(1) In or out character: in
is incoming (it moves
towards the horizon) while the three other modes
are outgoing (they move away from the horizon).

(2) Energy sign: out
carries a negative energy and a
negative norm, see Eq. (12). (We are considering the
complex conjugated so that out
is a positive norm
mode). The three other modes have positive energy
and norm.
(3) Co- or counterpropagating nature: co;out
is copro
pagating (its group velocity in the frame of the fluid
is positive) while the three others are counterpropagating. This separation is useful because only the
latter are significantly mixed in a transonic flow
[9,10]. In effect, co;out
acts essentially as a spectator.

For M < 0, one would describe a black-hole flow. The

forthcoming analysis applies by reversing the sign of velocities
and the in or out character of the modes, see [9].
2
In the supersonic region, only two real wave vectors remain:
co;in
k and kin
. They both describe incoming modes.

081503-3

RAPID COMMUNICATIONS

YVES AURGAN et al.

PHYSICAL REVIEW D 92, 081503(R) (2015)

To get the S matrix, we need to identify the basis of

out
incoming (outgoing) modes in
( ) which contain only
one incoming (respectively outgoing) asymptotic plane
wave. For < max, there are three modes, so the scattering matrix has a size 3 3, and is an element of the Lie
group U2; 1 since  has a negative norm [9]. In this
paper we focus on the incident mode in
for a white-hole
flow. It is a good candidate to probe the analog Hawking
effect, and was studied in hydrodynamic flows
[16,17,26,27]. For x , it is a sum of four asymptotic
modes
co;out
in
out
out 
in
:
A

15

In transonic flows, there is no transmitted wave [26]. in

thus vanishes for x . When working with asymptotic

modes of unit norm, the norm of in
evaluated at late time
(in the sense of a broad wave packet),
2
2
2
2
N out
j j j j jA j ;

16

must be exactly 1 because of the conservation of Eq. (11).

2
[In stationary flows, N out
1 also expresses the conservation of the wave energy, see Eq. (12), and that of the
energy flux of Mhring [28].] The minus sign in front of
j j2 is the signature of an anomalous scattering. It stems

from the negative norm carried by out
, see the above
point (2). The coefficient thus mixes modes of opposite
norms and energies. In quantum settings, j j2 would give
the mean number of spontaneously produced particles from
amplifying vacuum fluctuations, that is, the Hawking
radiation [2]. The gravitational analogy [1,4] indicates that
j j2 should follow a Planck law when dispersion effects
(and grey body factors [29,30]) are negligible. Moreover, it
predicts that the effective temperature T should be given by
T H =2, where is the surface gravity obtained from
the analog metric of Eq. (14). (We choose the units so that
kB = 1. T H and are thus both frequencies.) Using
Eq. (6), one gets
x cS jcS M

M3
:
bj
2 x cS M

parameters (given in the caption of Fig. 3), so that the flow

is near critical: M=cS b1 1.054 and M=cS b2 0.943.
Using these parameters, one has max 0.0053, and
0.019 of the same order as the dispersive frequency scale
b c2S evaluated at the horizon. This means that we worked
just outside the weak dispersive regime [11]. Yet, for
frequencies up to max , j j2 follows rather well the
2
=T H
Hawking prediction jH
1, that is, a
j 1=e
Planck law with T H given by =2, see Eq. (17). At
2
low frequency, the relative difference j j2 =jH
j 1 is
of order 20% (when we used db 3, the difference reduced
to about 0.3%, as expected since we were then in a weakly
dispersive regime). Moreover, we see that the coefficient
A involving the copropagating mode can be safely
neglected as jA j2 remains smaller that 0.1%. In order to
estimate the numerical errors, we also show the quantity
2
N out
1. As explained below Eq. (16), this quantity
would identically vanish in the absence of numerical errors.
In brief, the properties we obtain are in close agreement
with those found in other media [9,12,23,31].

17

V. SPECTRAL ANALYSIS
We numerically solved the set of coupled equations on
the 1 1-dimensional fields and , using the method
of [26] adapted to the present case. The results concerning
the incoming mode of Eq. (15) propagating in a transonic
flow described by Eq. (7) are shown in Fig. 3.3 We stopped
the integration for slightly below the critical frequency
max , where and A both vanish. We tuned the various
3

FIG. 3 (color online). Plot of TH j j2 (blue, solid) and

jA j2 (orange, dashed) of Eq. (15) as functions of the frequency.
The black dot-dashed curve shows TH j j2 for a Planck law at
the Hawking temperature T H =2. The parameters are
M 1=3, b1 9, b2 7, and db 1. The green dotted line
2
out 2
represents N out
1 where N is given in Eq. (16). Its
nonvanishing value quantifies the numerical errors.

When sending a localized wave packet on a white-hole

horizon, we observed at late times the formation of an undulation,
see [22].

VI. EXPERIMENTAL ASPECTS

The incident waves will be sent by a loudspeaker in a
rectangular channel, and the scattered waves will be
observed using two arrays of microphones, see Fig. 1.
The frequency range will be chosen to fulfill the lowfrequency hypothesis used in Eq. (3). A stationary flow
with a mean Mach number M 0.3 will be provided in the
channel. The compliant wall will be realized with a
honeycomb structure. Its height Hbx will vary in such
a way that the flow is transcritical, i.e., that M=cS crosses
unity at some x, something which has not yet been reached
in water tank experiments [16,17] aimed at detecting the
analog Hawking radiation. Another important advantage is
the possibility of sending the three types of incident waves,

081503-4

RAPID COMMUNICATIONS

SLOW SOUND IN A DUCT, EFFECTIVE TRANSONIC

and therefore to measure the nine scattering coefficients, in

both black-hole and white-hole flows. We suspect that
working in these conditions can produce turbulence effects
and whistling. To reduce these effects, the compliant wall
will be covered by a wire gauze with a very low flow
resistance. Special attention should also be devoted to
dissipation. In spite of these experimental difficulties, the
system seems to be a good candidate to probe the various
aspects of the analog Hawking radiation.

PHYSICAL REVIEW D 92, 081503(R) (2015)

reduction was performed, exhibiting an analog metric and

the first dispersion terms while retaining the Hamiltonian
structure. The Hawking spectrum was numerically recovered for sufficiently slowly varying profiles of the compliant wall. We thus hope to be able to verify that the norm
of the anomalous coefficient j j2 grows as =2 for
low frequency.
ACKNOWLEDGMENTS

We showed that a low-Mach-number uniform flow of air

in a tube with a compliant wall can produce a sonic horizon
by reducing the local effective one-dimensional speed
of sound. Despite the unusual boundary condition at the
compliant wall, the problem was phrased in a Hamiltonian
formalism. For near-critical flows, a (1 1)-dimensional

We thank Scott Robertson for his remarks on a preliminary version of this paper. P. F. is grateful to the LPT for its
hospitality during an internship sponsored by the cole
Normale Suprieure. We acknowledge partial support from
the French National Research Agency under the Program
Investing in the Future Grant No. ANR-11-IDEX-0003-02
associated with the project QEAGE (Quantum Effects in
Analogue Gravity Experiments).

[1] W. Unruh, Phys. Rev. Lett. 46, 1351 (1981).

[2] S. W. Hawking, Commun. Math. Phys. 43, 199 (1975).
[3] R. Brout, S. Massar, R. Parentani, and P. Spindel, Phys. Rep.
260, 329 (1995).
[4] W. Unruh, Phys. Rev. D 51, 2827 (1995).
[5] R. Brout, S. Massar, R. Parentani, and P. Spindel, Phys. Rev.
D 52, 4559 (1995).
[6] P. Horava, Phys. Rev. D 79, 084008 (2009).
[7] T. Jacobson, Phys. Rev. D 81, 101502 (2010).
[8] A. Coutant and R. Parentani, Phys. Rev. D 90, 121501
(2014).
[9] J. Macher and R. Parentani, Phys. Rev. D 79, 124008
(2009).
[10] A. Coutant, R. Parentani, and S. Finazzi, Phys. Rev. D 85,
024021 (2012).
[11] S. Finazzi and R. Parentani, Phys. Rev. D 85, 124027
(2012).
[12] S. J. Robertson, J. Phys. B 45, 163001 (2012).
[13] L. Garay, J. Anglin, J. Cirac, and P. Zoller, Phys. Rev. Lett.
85, 4643 (2000).
[14] R. Schutzhold and W. G. Unruh, Phys. Rev. D 66, 044019
(2002).
[15] T. G. Philbin, C. Kuklewicz, S. Robertson, S. Hill, F. Knig,
and U. Leonhardt, Science 319, 1367 (2008).
[16] G. Rousseaux, C. Mathis, P. Maissa, T. G. Philbin, and
U. Leonhardt, New J. Phys. 10, 053015 (2008).

[17] S. Weinfurtner, E. W. Tedford, M. C. Penrice, W. G. Unruh,

and G. A. Lawrence, Phys. Rev. Lett. 106, 021302 (2011).
[18] O. Lahav, A. Itah, A. Blumkin, C. Gordon, S. Rinott, A.
Zayats, and J. Steinhauer, Phys. Rev. Lett. 105, 240401
(2010).
[19] J. Steinhauer, Nat. Phys. 10, 864 (2014).
[20] M. K. Myers, J. Sound Vib. 71, 429 (1980).
[21] Y. Aurgan and V. Pagneux, J. Acoust. Soc. Am. 138, 605
(2015).
[22] See Supplemental Material at http://link.aps.org/
supplemental/10.1103/PhysRevD.92.081503 for a discussion of the wave equation and the formation of an undulation.
[23] J. Macher and R. Parentani, Phys. Rev. A 80, 043601 (2009).
[24] A. Coutant and R. Parentani, Phys. Fluids 26, 044106 (2014).
[25] C. Barcelo, S. Liberati, and M. Visser, Living Rev. Relativity 14, 3 (2011).
[26] F. Michel and R. Parentani, Phys. Rev. D 90, 044033
(2014).
[27] L.-P. Euv, F. Michel, R. Parentani, and G. Rousseaux,
Phys. Rev. D 91, 024020 (2015).
[28] W. Mhring, J. Fluid Mech. 431, 223 (2001).
[29] D. N. Page, Phys. Rev. D 13, 198 (1976).
[30] P. R. Anderson, R. Balbinot, A. Fabbri, and R. Parentani,
Phys. Rev. D 90, 104044 (2014).
[31] F. Michel and R. Parentani, Phys. Rev. D 88, 125012 (2013).

VII. CONCLUSIONS

081503-5